RU2627143C1 - Method of detecting low-contrast point objects - Google Patents

Abstract

FIELD: physics.

SUBSTANCE: method of registering low-contrast point objects uses an optoelectronic device, the lens of which forms optical images of point objects on the matrix radiation receiver. Images of point objects are exposed once and element-by-element read without pixel binning. The generated digital data array is copied in four instances and then processed each instance separately with the binary elements of the digital array of data. The first instance is processed in its original form, the second instance is processed from the second column of the digital array of data, the third instance is processed from the second line of the digital array of data, the fourth instance is processed from the second column and the second row of the digital array of data.

EFFECT: increasing the speed of information processing.

6 dwg, 5 tbl

Description

The invention relates to optical and optoelectronic technology and can be used to register low-contrast point objects, for example, artificial and natural celestial bodies.

Known methods for recording point objects, which include the formation of optical images of objects on the photosensitive surface of a multi-element array radiation detector, for example, a CCD matrix, by a single exposure during the selected time interval and by reading the accumulated charges. Optical images of point objects, which are scattering circles, in which useful photons are usually distributed according to the normal law, are projected onto specific elements - pixels of the CCD matrix. Each of these pixels is actually a separate converter of photons into photoelectrons, i.e. is an elementary radiation receiver. To obtain more complete information about objects, in particular to increase the accuracy of measuring their coordinates, the pixel sizes are chosen smaller than the scattering circles. In this case, useful photons, i.e. Photons from point objects are divided between illuminated pixels, that is, pixels that hit the image of a point object. As a result, for low-contrast objects, the signal-to-noise ratio at the output of one pixel is small and the probability of correct registration of objects is low.

To increase the probability of correct registration, instead of a single exposure and reading of charges, multiple exposure and reading are used, i.e. perform several registrations of the same objects. It is believed that with N-fold exposure and reading, i.e. with an increase in the number of registrations by N times, the sensitivity of the recording device increases by N ^1/2 times - this is described in the publication "Isolation of Optical Signals Against the Background of Random Noise" by N. Shestov - M .: Owls. Radio, 1967, p. 145. However, when objects are detected under time pressure conditions, restrictions are imposed on the number N, which reduces the effectiveness of registration.

There is also a known method of registering objects, which consists in using binning of the photosensitive elements of the CCD. When binning, charges are not read from each individual elementary pixel, but from an array of elements with specified sizes, for example 2 × 2 pixels, i.e. pixels are combined into groups and the accumulated charge is read. As a rule, independent binning is applied in two coordinates with coefficients 2, 3, 4, etc. With this binning, the charges accumulated in elementary pixels are combined, which increases the number of useful photoelectrons. So, for example, with an exposure time of 1 second, binning increased the penetration of the AZT-8 telescope from 16.5 ^m to 17.0 ^m , according to test experiments conducted on AZT-8 in Yevpatoriya, the author of the article is I. Molotov, link - lfVn. astronomer.ru/optic/evpatoria/exp2006/index.htm.

However, with this method of recording low-contrast point objects, for example, stars located randomly on the celestial sphere, these stars are not in equal conditions. If the image of a star over an area occupies four pixels of a matrix radiation detector, then this image can hit pixels in such a way that the charges from a faint star during binning are counted at a time. In this case, the summation of useful photoelectrons will occur and the signal-to-noise ratio in the read signal will be high, which will provide a more reliable registration of the star. But with the same probability, the image of a star can hit four pixels of the matrix radiation detector in such a way that, when binning, the addition of useful photoelectrons does not occur and the signal-to-noise ratio remains low, i.e. binning will not only not improve the conditions of registration, but will even worsen, since the number of interfering electrons will increase due to the fact that these electrons add up.

The closest analogue to the claimed technical solution is the method of registration of low-contrast point objects described in the patent of the Russian Federation for invention No. 2574522, IPC G06T 7/60, H04N 5/30, published 02/10/2016. To implement this method, an optical-electronic device is used, whose lens forms optical images of point objects on the photosensitive surface of the matrix radiation detector, which consists of many pixels forming the rows and columns of the matrix radiation detector. These images are repeatedly exposed and read with pixel binning, when adjacent pixels of the matrix radiation detector are grouped into one binable array, and with each new exposure and reading, the binable array is shifted along a row or column of the radiation matrix receiver with respect to the previous position. Due to such multiple displacements, randomly located stars appear to be in approximately equal conditions and an increase in the registration efficiency is provided for all stars.

However, multiple exposure and reading of the same plot increases the duration of the registration act. If the device operates in the view of the celestial sphere, this reduces the speed of view.

The task to which the invention is directed is to increase technical characteristics when registering low-contrast point objects.

The technical result is an increase in the speed of information processing and the speed of view of the celestial sphere.

This is achieved by the fact that in the method of recording low-contrast point objects, an optical-electronic device is used, the lens of which forms optical images of point objects on the photosensitive surface of the matrix radiation receiver, which consists of many pixels forming the columns and rows of the matrix radiation receiver, these images are exposed and read , in contrast to the known, in this method of registering low-contrast point objects, images of point objects are once exposed They are scanned and element-wise read without binning of pixels, the read electrons are stored in the form of a digital data array consisting of elements grouped into rows and columns, the format of which corresponds to the format of the matrix radiation receiver, the generated digital data array is copied in at least four copies, and then processed each instance separately with binning elements of a digital data array, the first instance being processed in its original form, the second instance being processed from the second column of digits A flat data array, the third instance is processed from the second row of the digital data array, and the fourth instance is processed from the second column and the second row of the digital data array.

In FIG. 1 shows the process of recording the first, second, third and fourth point objects during a single-element single reading of the accumulated charges without binning.

In FIG. Figure 2 shows the process of registering the first, second, third, and fourth point objects during a single exposure of the CCD matrix and reading the accumulated charges with binning, i.e. grouping four adjacent pixels of a matrix radiation detector into binable arrays of a matrix radiation detector.

In FIG. Figure 3 shows the process of recording the first, second, third, and fourth point objects during four-fold reading of accumulated charges with binning, when adjacent pixels of the matrix radiation detector are grouped into one binable array, and with each new exposure and reading, the binable array is shifted along the row or column of the radiation matrix detector in relation to the previous position - a, b, c, d.

In FIG. Figure 4 presents a fragment of a digital data array in its original form, consisting of rows and columns of elements containing digital information on the total number of electrons, including photoelectrons from a point object, accumulated and read from the corresponding pixels of the radiation matrix receiver, as well as interfering electrons generated during the accumulation and charge reading.

In FIG. 5 presents four copies of the digital data array, and the processing options for each of them are a, b, c, d. Namely, the first copy is processed in its original form (Fig. 5a). The second instance is processed from the second column of the digital data array, the third instance is processed from the second row of the digital data array, and the fourth instance is processed from the second column and the second row of the digital data array.

In FIG. Figure 6 shows the result of binning four copies of a digital data array - a, b, c, d.

The figures indicate: image 1 of the first point object, image 2 of the second point object, image 3 of the third point object, image 4 of the fourth point object, pixels 5 of the radiation matrix receiver, binable arrays 6 of the radiation matrix receiver, elements 7 of the digital data array containing digital information only on the number of interfering electrons, elements 8 of a digital data array containing digital information on the total number of photoelectrons from the first point object and interference x electrons, elements 9 of a digital data array containing digital information on the total number of photoelectrons from the second point object and interfering electrons, elements 10 of a digital data array containing digital information on the total number of photoelectrons from the third point object and interfering electrons, elements 11 of a digital data array containing digital information about the total number of photoelectrons from the fourth point object and interfering electrons.

It is assumed that the size of the images of point objects is equal to the size of four pixels of the matrix radiation detector. For example, the diameter of the image 1 of the first, the diameter of the image 2 of the second, the diameter of the image 3 of the third and the diameter of the image 4 of the fourth point objects is 0.024 mm, the pixel sizes 5 of the matrix radiation detector are 0.012 × 0.012 mm ² , the dimensions of the binable arrays 6 of the matrix radiation receiver are 0.024 × 0.024 mm ² . All point objects have approximately the same brightness and are low-contrast. Due to the random distribution of the images over the photosensitive surface of the matrix radiation detector, the image of a particular point object can get at four pixels, at worst, at nine pixels of the receiver. In this case, useful photoelectrons are divided between the pixels.

Consider the process of registering low-contrast point objects using the proposed method using an optoelectronic device, such as an astronomical device. The registration process of low-contrast point objects is implemented by an astronomical device as follows. The field of view of an optoelectronic device containing a lens, a matrix radiation detector, which consists of many pixels forming rows and columns, an analog-to-digital converter, a computing unit, and recording process control equipment, is displayed on the studied area of the celestial sphere and Earth rotation is turned on. Images of recorded point objects and a distributed sky background are projected onto specific pixels 5 of the matrix radiation detector, as shown, for example, in FIG. 1, and remain practically stationary during the selected exposure time. When exposed, useful photons from objects and interfering photons from a distributed sky background are converted into photoelectrons and accumulate in pixels 5 of the matrix radiation detector. In the process of accumulation, thermoelectrons and other interfering electrons formed in the matrix radiation detector are also added to them.

Next, the process of a single element-wise reading of the accumulated charges and their analog-to-digital conversion is performed. Based on the results of reading and analog-to-digital conversion, a digital data array is formed in the computing unit, consisting of rows and columns, the format of which corresponds to the format of a matrix radiation receiver (Fig. 4). An array of data is a uniform, ordered, structured data type with direct access to elements. The elements of the array are united by a common name and occupy a certain finite memory area in the computing unit. Any element of the array can be accessed by specifying the name of the array and the index of the element in the array.

After the completion of the formation of the digital data array, the transfer of the field of view of the optoelectronic device to the next section of the celestial sphere begins. During this transfer, the digital data array is copied in at least four copies and each instance is further processed separately (Fig. 5). The first copy is processed in its original form, without changes (Fig. 5A). Processing of the second instance begins with the second column of the digital data array without capturing the first column (Fig. 5b). The third instance is processed from the second row of the digital data array without capturing the first row (Fig. 5c). The fourth instance is processed from the second column and second row of the digital data array without capturing the first column and the first row (Fig. 5d). The processing results of each of the four instances of the digital data array are shown in FIG. 6a-6g, respectively.

When using a large-format matrix radiation receiver, a single reading of each of the instances of the digital data array can be performed in parallel so that the duration of the process does not exceed the time of transfer of the field of view to an adjacent portion of the celestial sphere. This parallelization of the process of reading digital elements of a digital data array with binning allows you to provide a high rate of view of the celestial sphere in almost any hardware implementations of the proposed method.

Consider the quantitative characteristics of the known methods for registering point objects and the proposed method. The speed V _{about the} observation of the celestial sphere by an optical-electronic device is measured, as a rule, in square degrees / h and can be found by the approximate formula:

V _rev = 3600⋅S _pz / (t _e + t _c + t _per ),

where S _pz - field of view of the device in square. hail.;

t _e - exposure time in s;

t _c is the charge reading time in s;

t _lane is the time of transfer of the field of view to the neighboring section in s.

For specific calculations, the following characteristics are accepted:

S _pz = 1 sq. hail.;

t _e = 1.0 s;

t _c = 0.1 s;

t _lane = 1.5 s.

Then, for the method of recording point objects during a single exposure of the CCD matrix and reading the accumulated charges with binning the pixels of the radiation matrix detector (Fig. 2), we obtain:

V _about = 3600⋅1 / (1.0 + 0.1 + 1.5) = 3600 / 2.6 = 1385 sq. city / h

For the method of registering point objects with four times reading the accumulated charges with binning, when with each new exposure and reading the binable array is shifted along the row or column of the radiation matrix receiver with respect to the previous position (Fig. 3), we obtain:

V _about = 3600⋅1 / (4⋅1.0 + 4⋅0.1 + 1.5) = 3600 / 5.9 = 606 sq. city / h

For the proposed method, when digital elements are binded separately for each instance of a digital data array, we obtain:

V _about = 3600⋅1 / (1.0 + 0.1 + 1.5) = 3600 / 2.6 = 1385 sq. city / h

The probabilities of correct registration point objects P _ave at a probability of false registration ≤0,001 F _np for the method of FIG. 2 are indicated in table. 1 (the calculation is given in the patent of the Russian Federation No. 2574522).

The probabilities of correct registration point objects P _ave at a probability of false registration ≤0,001 F _np for the method of FIG. 3 (the closest analogue) are indicated in table. 2 (the calculation is given in the patent of the Russian Federation No. 2574522).

We estimate the probability of the correct registration point objects P _ave at a probability of false registration ≤0,001 F _np for the inventive method.

It is believed that the flux of photons entering the input of the device is described by Poisson statistics. Then the average total charge read from the pixel of the matrix radiation detector contains n _eo photoelectrons from the object and n _ep electrons, which are interference. For n _{en the} expression is true:

n _en = n _ef + n _es

where n _ef is the number of photoelectrons from the sky background generated in one pixel during the exposure time t _e ,

n _es - the number of noise electrons generated in one pixel during the exposure time t _e and the signal read time t _c .

In turn, for the internal noise of the equipment n _es can be written:

n _ew = n _a + n _ec,

where n _em - thermoelectric noise in electrons per second from one pixel during exposure,

n _ec - reading noise in electrons from one pixel in one act of reading.

Then n _en = n _ef + n _em + n _ec .

To evaluate n _eo and n _ef , approximate formulas given in the book by A. Abramenko can be used. et al. Television Astronomy, ed. Nikonova V.B. M .: Nauka, 1974, p. 156, formulas (8.18) and (8.19):

N _eo = 3.8⋅10 ^(4-0.4⋅mo) ⋅ε⋅D ² ⋅τ _о ⋅τ _а ⋅t _e ,

N _eph = 1,3⋅10 ^{(15-0,4⋅mf)} ⋅ε⋅ (D / ^{f) 2 2} ⋅d ⋅τ _about ⋅t _e.

where N _eo is the number of useful photoelectrons formed in those parts of the pixels of the CCD matrix on which the image of the object fell,

N _ef is the number of photoelectrons from the sky background formed in those parts of the pixels of the CCD matrix, on which the image of the object fell,

mо is the brightness of the recorded point object in magnitude,

mf is the level of the distributed background in magnitude with sq. angle sec.,

ε is the integral quantum output of the radiation receiver,

D is the diameter of the effective entrance pupil of the lens in mm,

f is the focal length of the lens in mm,

d is the diameter of the image of the recorded point object in mm,

τ _{about the} integrated transmittance of the optics,

τ _a is the integral transmittance of the atmosphere,

t _e - exposure time in seconds.

From FIG. 1 it follows that the image 1 of the first point object, the image 2 of the second point object, the image 4 of the fourth point object were distributed equally between four pixels, i.e. 25% of the useful photoelectrons N _eo accounted for each of the illuminated pixels. Then for one pixel the relation is true:

n _eo = 0.25⋅N _eo .

Image 3 of the third point object was distributed in the following proportions: two pixels each accounted for 24% of the photoelectrons N _eo and another four pixels each accounted for 13% of the photoelectrons N _eo . Then for one illuminated pixel the following relations are true:

n _eo = 0.24⋅N _eo ,

n _eo = 0.13⋅N _eo .

The value of N _eff corresponds to the number of photoelectrons from the sky background, formed on the area of the photosensitive surface of the matrix radiation detector, equal to the image area of a point object. Then for one pixel the relation is true:

For specific calculations, the values of the main characteristics of the astronomical instrument are taken:

D = 180 mm;

f = 1500 mm;

d = 0.024 mm;

τ _about = 0.65;

ε = 0.7 photoelectric / photon;

t _e = 1.0 s;

l _p = 0.012 mm;

n _em = 2 electric / s / pix;

n _ec = 11 electr./pix./act count

It is assumed that the brightness of the recorded point objects in magnitude corresponds to 17.5 ^m . The following values are accepted for the conditions for registering point features:

atmospheric transparency τ _a = 0.7;

the level of the distributed sky background mf = 21 stars. Vel./sq. angle sec .;

the registration threshold γ is chosen so that the condition P _lr ≤0.001 is fulfilled.

For given characteristics, the number of useful photoelectrons per image of a point object:

N _eo = 3.8⋅10 ^{(4-0.4⋅17.5)} 0.7⋅180 ² ⋅0.65⋅0.7⋅1.0 = 39.2.

The number of photoelectrons from a point object per one specific pixel of the receiver and, accordingly, one specific element of a digital data array:

n _eo = 0.25⋅N _eo = 0.25⋅39.2 = 10;

n _eo = 0.24⋅N _eo = 0.24⋅39.2 = 9;

n _eo = 0.13⋅N _eo = 0.13⋅39.2 = 5.

The number of photoelectrons from the sky background per square area on the image of a point object with a diameter of d = 0.24 mm:

N _eph ^{= 1,3⋅10 (15-0,4⋅21) ⋅0,7⋅ (180/1500} ) ^{2 2} ⋅0,024 ⋅0,65⋅1,0 = 19,53.

The number of photoelectrons from the sky background per one pixel of a matrix radiation detector of size l _p = 0.12 mm:

n = 19,53⋅0,012 _eph ² / (π⋅0,024 ^2/4) = 6.

The number of interfering electrons per pixel of the matrix radiation detector and, accordingly, one element of a digital data array:

n _en = 6 + 2 + 11 = 19.

The probability of the correct registration of the object R _pr and the probability of false registration of the object R _lr during multiple reading of the elements of the digital data array using binning can be represented by approximate formulas given in the book by Van Tris G. “Theory of detection, estimation and modulation”. Volume 1, M., “Sov. Radio ”, 1972, p. 53, formulas (84) and (85):

where n _eo∑ is the total number of photoelectrons from a point object in binable elements,

n _ep∑ is the total number of interfering electrons in binable elements.

When repeatedly reading the elements of a digital data array using binning, the detection threshold γ is selected by the Neumann-Pearson criterion. The criterion is that the value of the probability P _lr is set in advance, the probability P _{pr is} maximized.

For total interference electrons during binning, we can write:

n _en∑ = n _en + n _en + n _en + n _en = 19 + 19 + 19 + 19 = 76.

Calculations show that for a given level of interference and for γ = 98 the value of P _lr <0.001. Consequently, the condition is satisfied, and further calculations _{P, etc.} are carried out with γ = 98.

With a single reading of instances of a digital data array, their digital elements containing information about photoelectrons from a point object and from interfering electrons can be combined into binable arrays in various ways. These options are presented in FIG. 5 and FIG. 6.

A binable array of a digital data array includes information about one element containing 25% of the photoelectrons from a point object and interfering electrons and about three elements with interfering electrons. The total number of electrons in the binning array:

n _eo∑ + n _ep∑ = 29 + 19 + 19 + 19 = 86.

For a binable array consisting of 86 electrons (10 photoelectrons from a point object and 76 interfering electrons), the probability of correct registration of a point object will be:

A binable array of a digital data array includes information about two elements containing 25% of photoelectrons from a point object and about interfering electrons (only two elements of 50% of photoelectrons from a point object) and about two elements with interfering electrons. The total number of electrons in the binning array:

n _eo∑ + n _ep∑ = 29 + 29 + 19 + 19 = 96.

For a binable array consisting of 96 electrons (20 photoelectrons from a point object and 76 interfering electrons), the probability of correct registration of a point object will be:

A binable array of a digital data array includes information on four elements containing 25% of photoelectrons from a point object and on interference electrons (only four elements contain 100% of photoelectrons from a point object). The total number of electrons in the binning array:

n _eo∑ + n _ep∑ = 29 + 29 + 29 + 29 = 116.

For a binable array consisting of 116 electrons (40 photoelectrons from a point object and 76 interfering electrons), the probability of correct registration of a point object will be:

A binable array of a digital data array includes information about one element containing 13% of the photoelectrons from a point object and about interfering electrons and about three elements with interfering electrons. The total number of electrons in the binning array:

n _eo∑ + n _ep∑ = 24 + 19 + 19 + 19 = 81.

For a binable array consisting of 81 electrons (5 photoelectrons from a point object and 76 interfering electrons), the probability of correct registration of a point object will be:

A binable array of a digital data array includes information about one element containing 13% of photoelectrons from a point object and from interfering electrons, about one element containing 24% of photoelectrons from a point object and from interfering electrons (only two elements 37% of a point object) and about two elements with interfering electrons. The total number of electrons in the binning array:

n _eo∑ + n _ep∑ = 24 + 28 + 19 + 19 = 90.

For a binable array consisting of 90 electrons (14 photoelectrons from a point object and 76 interfering electrons), the probability of correct registration of a point object will be:

A binable array of a digital data array includes information on two elements containing 13% of photoelectrons from a point object and from interfering electrons (only two elements contain 26% of photoelectrons from a point object) and about two elements with interfering electrons. The total number of electrons in the binning array:

n _eo∑ + n _ep∑ = 24 + 24 + 19 + 19 = 86.

For a binable array consisting of 86 electrons (10 photoelectrons from a point object and 76 interfering electrons), the probability of correct registration of a point object will be:

A binable array of a digital data array includes information on two elements containing 13% of the photoelectrons from the object and jamming electrons and two elements containing 24% of the photoelectrons from a point object and from interfering electrons (only four elements contain 74% of photoelectrons from a point object ) The total number of electrons in the binning array:

n _eo∑ + n _ep∑ = 24 + 24 + 28 + 28 = 104.

For a binable array consisting of 104 electrons (28 photoelectrons from a point object and 76 interfering electrons), the probability of correct registration of a point object will be:

Probability calculation results correct registration point objects P _ave for embodiments in association biniruemye arrays of digital data array elements shown in Table. 3.

We determine the probability of the correct registration of the first, second, third and fourth point features when using the proposed method. For this we use the formulas for addition and multiplication of probabilities, as well as the data given in table. 3 and shown in FIG. 5, FIG. 6.

The probability of correct registration P _{pr of the} first, second and fourth point objects during a single reading of instances of a digital data array with binning of elements:

P _ave = 1- (1-0,00973) ⋅ (1-0,25826 ) ⋅ (1-0,00973) ⋅ (1-0,25826) ⋅ (1-0,94677) ⋅ (1-0, 25826) ⋅ (1-0.00973) ⋅ (1-0.25826) ⋅ (1-0.00973) =

= 1-0.99027⋅0.74174⋅0.99027⋅0.74174⋅0.05323⋅0.74174⋅0.99027⋅0.74174⋅0.99027 =

= 1-0.01549 = 0.98451.

The probability of correct registration P _{pr of the} third point object when reading instances of a digital data array with binning of elements:

P _ave = 1- (1-0,00122) ⋅ (1-0,03163 ) ⋅ (1-0,03163) ⋅ (1-0,00122) ⋅ (1-0,00973) ⋅

· (1-0.64889) ⋅ (1-0.64889) ⋅ (1-0.00973) ⋅ (1-0.00122) ⋅ (1-0.03163) ⋅ (1-0.03163) ⋅

· (1-0.00122) = 1-0.99878⋅0.96837⋅0.96837⋅0.99878 0.99027⋅0.35111

⋅0.35111⋅0.99027⋅0.99878⋅0.96837⋅0.96837⋅0.99878 = 1-0.10579 = 0.89421.

The probability that after completing a single reading of instances of a digital data array with binning of elements, all four point objects will be registered:

P _ave = 0,98451⋅0,98451⋅0,89421⋅0,98451 = 0,85320.

The evaluation results are given in table. four.

In the table. 5 shows the viewing speed of the celestial sphere and the probability of detecting point objects for the methods shown in FIG. 2, FIG. 3, FIG. 5.

The proposed method has a high probability of correct registration of all four point objects and at the same time, a high speed of viewing the celestial sphere.

Thus, the application of the proposed method for registering low-contrast point objects makes it possible to achieve the indicated technical result, namely, to increase the speed of viewing the celestial sphere and the processing speed of information when registering low-contrast point objects.

Claims (1)

A method for recording low-contrast point objects, which consists in using an optical-electronic device, the lens of which forms optical images of point objects on the photosensitive surface of a matrix radiation detector, which consists of a plurality of pixels forming columns and rows of a matrix radiation detector, these images are exposed and read, characterized in that in this method of registering low-contrast point objects, images of point objects are once exposed and pixels are read out without binning, the read electrons are stored in the form of a digital data array consisting of elements grouped into rows and columns, the format of which corresponds to the format of the matrix radiation receiver, the generated digital data array is copied in at least four copies, and then each instance is processed separately with binning the elements of the digital data array, the first instance being processed in its original form, the second instance being processed from the second column of the digital m data assortment, the third instance is processed from the second row of the digital data array, and the fourth instance is processed from the second column and the second row of the digital data array.

Images